Adaptive Neural Network Control of Zero-Speed Vessel Fin Stabilizer Based on Command Filter

Round 1

Reviewer 1 Report

Dear authors,
Quite an interesting article.

1. The article lacks a description of the models used so far in Neural networks related to the issues contained in the article,
   2. Justify the choice of the model used in comparison to other Neural Networks models.
   3. Was Matlab or specialized software used in the research work? What other neural network models could be used?
   4. How was the movement of the object simulated? what software? Matlab?
   5. indicate the purpose of the work, why should such a model need to be developed, what kind of innovation does the absorbed model contain compared to other models?
   6. Formula (26) is it an error or a precision?
   7. Chapter 4. L 98m, W 10.2 m. In my opiinion displacement should be larger. more than 2000-2500 T. Draft, 4.5 - 5m. It all depends on the fullness of the hull. What is the full body hull coefficient in the proposed model?
   8. Conclusion should be more elaborate. Too meager conclusions. The discussion chapter is missing.

Author Response

See upload file for detailed reply

Author Response File: Author Response.pdf

Reviewer 2 Report

General comments and remarks:

The paper addresses a narrow range of a selected marine engineering problem, which is the application of fin stabilizers at zero speed when the vessel is at anchor. The paper has two dimensions. The first is solving a control system problem by using a command filter and combining Auxiliary Dynamic System, Nonlinear Disturbance Observer, and Adaptive ANN with including uncertainty and unknown disturbance. The designed control system has been checked properly for stability, too. The second is analyzing the application of the conducted control system for a selected vessel only by simulation, and comparing the results with the adaptive ANN control strategy but without a command filter. The first is well formulated and solved. The second, however, is not worked out and analyzed, sufficiently. In details:

1. Usually, fin stabilizers are used for small or medium-size vessels with high speed, since the effectiveness of fins directly depends on the vessel speed to the power of two. Zero speed fin stabilizing in literature does not refer to exactly zero speed, but low speeds down to zero. Therefore, it addresses a range of velocity not only speed of vessel equal zero. For other velocities rather than zero, the forces include also hydrodynamic ones. Additionally, when the ship is at anchor, depending on the sea condition the anchor chain forces influence the ship roll. These forces are completely omitted in the paper (see Eqs. (7)-(9)).
2. It is mentioned that one of the main features of the paper is delivering a solution applicable for situations in which large roll angles occurred. In fact, the high sea states for those sea areas that are usually selected for anchoring are observed rarely. If a large roll angle has occurred, mainly it is related to the resonance phenomenon. Therefore, it is very important to take into account the wave and ship’s parameters to identify the resonance frequency range. This is addressed by many authors in the related literature but is not considered and discussed in the paper, at all. Additionally, there is no justification why the existing solutions are unsatisfactory for large rolls –as is mentioned on Page 1.
3. The solution is checked only by simulation. The validation is done only by comparing the results with the simulation results when an adaptive ANN control strategy without a command filter is applied. Without experimental results, either for the ship model or the full-scale ship, the validation of the results remains under question. Please see the following article (not mentioned in the references), as an example of the correct validation process:

Lihua Liang, Peng Zhao, et al, Simulation and experimental study on the control strategy of zero-speed fin stabilizer based on disturbance and compensation, October 2018, PLoS ONE 13(10):e0204446, DOI: 10.1371/journal.pone.0204446.

Even comparing the results with the case in which a conventional PID is used, would enhance the current level of validation of the results.

Other comments:

1. Please use the professional nomenclature of Naval Architecture and Marine Engineering. For example “the length between the two columns” (Pages 3 and 12) should be “the length between two perpendiculars”; “… the initial stability height of the vessel…” should be “…the initial metacentric height of the vessel…”; “… is the stability arm …” (Page 3) should be “… is the righting arm …”, etc.
2. Justify Eqs. (25), (32), (37), and (39). Please deliver a description of the reason for selecting these relationships.
3. Please comment on the values of estimated control gain presented in Fig. 1f and justify why it is zero from the beginning of the simulation.
4. The learning rate of ANN firstly has been mentioned as “0.00001 I” (Page 12, Line 186) and then as “0.001 I” (Page 12, Line 191). Please make the necessary correction.
5. “a_e” in Eq. (10) is not defined.
6. Language and typos: A proofreading is strongly recommended. Some examples are mentioned hereafter.

• Please use a single space between every two successive words. This rule is not regarded, frequently.
• 1 is divided into 6 parts identified by letters “a” to “f”, but these parts are referred to as Figs. 1 to 6 on Page 12 (Lines 194 to 207).
• “structure of the” is repeated on Page 4: “… according to factors such as the structure of the structure of the vessel’s …”.
• On Page 6: “Using (25), the error variable is defined …” should be “Using (25), the error variables re defined …”
• On Page 7: “According to (24) and (26) …” should be “According to (24), (26) and (28) …”

 

Conclusion:

The designed control system and related analysis are valuable. However, to show the results are meaningful in practice the aspects of marine engineering practice should be included in the paper. A major revision is recommended.

Author Response

See upload file for detailed reply

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Dear Authors of the Manuscript,
Thanks for the detailed answers. I would only ask which type of ship - frigate was taken into account for the model tests.
Besides, I have no objections to the article. Great job.

Author Response

For the details responses, please see the response letter.

Author Response File: Author Response.pdf

Reviewer 2 Report

Thank you for the revised version, as well as your answers to previous comments. 

The paper still needs more clarifications and explanations, as described below:

1. Re Comment 1: Please clearly write in the paper that when the ship is at anchor, depending on the sea condition the anchor chain forces influence the ship roll, and also the hydrodynamic forces are omitted in Eqs. (7)-(9), so the model is incomplete. However, these forces can be regarded as unknown dynamics, which can be reconstructed by the adaptive neural network and the disturbance observer. The reader should be aware that your model is not appropriate for other cases.
2. Re Comment 2: Please clearly write that your study delivers only a control system solution and does not include practical application of fin stabilizers in which, as an example, the system should be carefully adjusted and analysed with respect to the resonance frequency. 
3. Re Comment 3: Please clearly write that the results have not been verified, yet. Please add the comparison of the response of the system, when a PID controller is applied to the paper (the same as you have written in your response to the review).
4. Re Comment 5: Please add the same justification you mentioned in the response to the review, to the paper.

Author Response

For the details responses, please see the response letter.

Author Response File: Author Response.pdf